In recent years, financial crises broke out frequently and showed contagion, which caused high attention from scholars both at home and abroad. This paper selects the appropriate dynamic conditional correlation model to study conditional correlation among the Eurozone, the American, A-share, and the Hong Kong stock market. By using the endogenous multiple structural break model and T-test method, the paper divides the contagion period and the stable period, and uses CCK model which considers the external influences to study herd behavior on the A-share market and the Hong Kong stock market. And then, this paper introduces the cross-sectional absolute deviation index of returns to study the herd behavior channel during two crises. It can be concluded that the spread of two crises' contagion to the Hong Kong stock market was faster than that to the A-share market, and the contagion period of the Hong Kong stock market lasted longer than that of the A-share market, but the contagion extent of the Hong Kong stock market was weaker than that of the A-share market. In addition, the contagion effect of the subprime crisis was stronger than that of the European debt crisis, but the contagion period of the subprime crisis lasted shorter than that of the European debt crisis. It confirms that herd behavior channel is one of the contagion channel to the A-share market and the Hong Kong stock market during two crises.
With the development of economic globalization, the fluctuations of commodity futures prices are increasingly violent. The uncertainty of the global economic situation and monetary policy make the commodity futures prices difficult to be accurately predicted. In this paper, corn, gold and crude oil are selected as the representatives of agricultural futures, industrial metals futures and energy futures, respectively. Based on the singular spectrum analysis (SSA) method, we decompose the commodity futures price and incorporate Kmeans dynamic clustering technique to group the decomposition series. Then the forecasting models of commodity futures price are developed by using extreme learning machine (ELM). The empirical results show that the decomposition and clustering schemes can significantly improve the accuracy of price forecasting, and performs well on the overall estimation and direction of change of the commodity prices.
In recent years, more and more researchers focused on how to use investor's view to assist portfolio selection. Black-Litterman model has been widely used because of its characteristics of combine investor's views and market equilibrium by Bayesian methods. This paper attempts to expend the Black-Litterman model to multi-period. Firstly, this paper solves the problem that the Black-Litterman model can't deal with views of extreme confidence. Then the paper establishes a new confidence matrix to expend the Black-Litterman model to multi-period framework. Finally, based on multi-period Black-Litterman model, this paper establishes a portfolio model with CVaR and empirically examines the model with mean-variance model and equal-weighted model as benchmark. The empirical results show that the multi-period Black-Litterman model performs better.
We develop a dynamic agency model which is based on Poisson jump and manager's risk-taking. Firstly, we consider an optimal contract design problem that assumes unobservable effort and risk-taking of agent. Equilibrium evolution of agent's value function is provided by using martingale method. Then we provide sufficient and necessary conditions for incentive compatibility contract. Furthermore, this paper gets the differential equation for the enterprise value where the optimal dynamic investment strategies can be obtained at the same time. Numerical analysis shows that, the opportunity to take risk has negative impact on the equity value and optimal investment. Taking into account punishment cost, shareholders should allow a manager's risk-taking behavior within a certain extent. Finally, the relationships between level of risk-taking and related coefficients are also discussed.